LLMs are fine. they're useful. they can generate code, and they can generate porn, and that's about 90% of what most people want. (mostly porn, for me, but that's because i enjoy writing my own code.)

they're shit because all the useful ones are owned by large tech corporations and, like any capitalists, they will fuck everything up, enshittify it, burn down the planet to build datacentres, whatever. obviously that's all bad. like i said the other day, corporations are cunts.

so we should be opposed to LLMs for the same reason we're opposed to corporations.

but this means it's fairly absurd to point at some code written by an LLM and say "haha, you used an LLM to write this? what an idiot!". i don't use LLMs to write code, but i have evaluated them for that, and yes, they sometimes produce code which is obviously or subtly wrong, but they can also produce exactly the same code i would have written myself.

pointing at a commit and saying "haha, this was written by an LLM, how silly!" makes you look like an idiot. LLMs can write useful code; that's an objective fact. if you're opposed to LLMs, you should oppose them for actual reasons, not for some reason you just made up.

if we want to get all moralistic about this, my main objection to LLMs is that writing code should be fun for its own sake, so why would you ever want an LLM to write your code? don't you like fun?

the only reason to do that is because you're a wage slave at some corporate job and your boss wants you to be more efficient, and there you are slaving away, and using an LLM makes your boss happy.

but what's the real problem there? it's not the LLM, it's your boss. your boss is a cunt. corporations are cunts. the problem is wage slavery, not LLMs.

if we got rid of wage slavery, we'd probably also get rid of LLMs, but that doesn't mean LLMs are the root cause of this problem.

Is AI the next Dumbwaiter?

The demand for AI's unbridled growth is reactionary — a way of doubling down on the same old colonizing way of doing things. It doesn't have to be.

Rushkoff

@lw honestly, Drew Devault put it in the best way:

https://drewdevault.com/2026/03/25/2026-03-25-Forking-vim.html

A eulogy for Vim

@omar i don't disagree with his logic here, but i disagree with his conclusions. i agree LLMs, as they currently exist, are bad, but that doesn't mean the *output* of LLMs is the fruit of the poisoned tree; if the output already exists, we may as well use, otherwise we are only harming ourselves, not the LLM providers.

i do understand the counter-argument (along the lines of, if we refuse any LLM-generated code, we will encourage people not to use LLMs) and i'm sympathetic to that, but ultimately i have not found this sort of moralistic approach to be useful in either generating actual change, or in improving the reality of the world we live in.

in other words: the LLM shit is happening whether we like it or not, so we might as well benefit from it where we can.